demo-day
Prepare a feature-review demo for a demo day — a story planned with the user, demo script, setup scenarios that seed the environment and print test data and users, a rehearsal recording each feature's demonstrability, and polished demo videos.
- License 许可
- MIT
- 所属仓库
- BetoFrega/skills
- 最近更新
- 2026-11-14
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- demo-day
- description
- Prepare a feature-review demo for a demo day — a story planned with the user, demo script, setup scenarios that seed the environment and print test data and users, a rehearsal recording each feature's demonstrability, and polished demo videos.
02 资源文件 5
- 参考文档agents/openai.yaml147 B
- 资产文件assets/demo-script-template.md1.1 KB
- 参考文档references/recording.md5.9 KB
- 脚本scripts/finish-recording.mjs5.7 KB
- 脚本scripts/recorder.mjs42.4 KB
03 SKILL.md 正文
Demo day
Read beto-mode first and apply the principles it routes to; name each principle that changed a choice. Global rules keep authority over decisions and permissions.
The invocation authorizes the demo artifacts below: scenarios, setup, demo script, rehearsal report, and demo videos. Product code, shared environments, and tracker writes beyond the verdict comment need their own authority.
1. Select features
The user names the features. For each, find its work item or PR and read the draft demo script from its validation section (see plan-increment), acceptance, and references. Without a draft, reconstruct the steps from the work item, PR, and code, and mark them as inferred per Separate Documented From Inferred.
Resolve the records location: an existing project convention or configured product
documentation; otherwise ask, defaulting to docs/demos/<YYYY-MM-DD>/. Record the
chosen convention in the project's agent-skills entry so later demo days reuse it.
2. Plan the story with the user
Before building anything, agree on the demo's story with the user through grilling. Open with one drafted proposal they can correct, one line per item, drafted from what step 1 read. Mark each guess. Guess where a wrong guess is cheap to correct (time budget, cast); ask where the answer would reshape the proposal (usually the audience). Put the first such question after the proposal, and continue the remaining ones as grilling rounds. The proposal is grilling's opening plan; its topics, progress counts, and session tree cover only the questions that stay open. The proposal covers:
- Audience: who watches, what they should leave believing, and the time budget. It sets how much theatre fits.
- Per feature: the problem and before/after; the key moment, the screen the audience came to see; the cut, which steps appear and which become starting state; the cast and data, the people, companies, and numbers that tell the story.
- Order of the features and how one leads into the next, when there are several.
Name what step 1 found that limits the story, such as a "before" that can no longer be shown, and how the cut handles it. A suspected product defect waits for step 5's classification.
Record the settled answers in the demo script's Audience section and each feature's Key moment and Cast lines. The scenarios in step 3 seed exactly that cast and data.
3. Build setup scenarios
Target the local environment. Use staging or another shared environment only when the user names it for this demo day; then isolate demo data under a per-demo marker, provide a cleanup command, and keep credentials out of tracked files and output records (Security by Design).
Give each feature a reusable, idempotent scenario, versioned in the project and built on its existing seed, factories, or fixtures. Extend that tooling before adding a new mechanism; when the project has no seed entry point, add the smallest one the scenarios need. Compose the demo day's setup from the scenarios: one command that brings the environment to the demo state and prints what the presenter needs — URLs, test users and credentials, record identifiers, and the state each feature starts in.
4. Write the demo script
Fill the demo script template per feature: the problem, the before/after, the observable steps on the audience-facing interface, what each step shows, and the reset or recovery for a step that goes wrong. Keep the order, cut, and key moments agreed in step 2, and estimate each feature's time.
5. Rehearse
Run the setup from a clean state, then perform every step through the interface the audience will see: drive UI with Playwright and capture a screenshot per step; use a CLI or API only when that is the feature's own interface. When the audience-facing interface cannot be driven here, the feature is not demonstrable; record exactly what blocked it.
On a failed step, classify the cause per Fix Root Causes:
- Demo artifact (scenario, setup, script): fix it and rehearse the feature again from a clean setup.
- Product defect: stop that feature, mark it not demonstrable, record the diagnosis and evidence, and bring a proposed fix to the user. Change product code only after their decision.
Assign each feature a verdict — demonstrable, demonstrable with caveats, or not demonstrable — supported by the rehearsal evidence (Evidence Before Claims).
6. Record videos
When the user asks for videos, record each demonstrable feature after its rehearsal passes, following recording.
- Plan the staging. Propose each step's staging over its rehearsal screenshot, in the same proposal-then-questions shape as step 2: captions and their text, which moment gets spotlight, zoom, or suspense, the motion style, chapter titles, and the cursor's acting: its entrance and exit and the expression at each turn of the story. Be bold on the key moment and brisk elsewhere. Record the answers in the Staging column.
- Record from a clean setup, following the Staging column, and check the frames.
- Get approval. Send each video and ask for direction notes. Turn each note into a Staging change and record that feature again. A video is done only when the user approves it (Evidence Before Claims); until then the report marks it awaiting approval.
7. Record and report
Save the filled template in the demo day's folder as the demo script and rehearsal report: each step's result, evidence links, timing, verdict, and video link with its approval state. Record only the verdict and a link to the report in each feature's canonical work item or record (One Canonical Record); features known only by PR keep the report alone.
Complete when every named feature has a verdict backed by a rehearsal from a clean setup, the setup command reproduces the demo state, each requested video is approved by the user, and each pending product decision is presented to the user. Report the verdicts, the setup command and its output, the pending decisions, and the delivery state of the versioned scenarios and records.
04 同仓库更多 Skills
writing-for-humans
Load before writing any reader-facing text, including messages, and follow its rules as you write. Also use to rewrite or audit existing prose for plain language, clarity, readability, or actionability.
writing-for-agents
Write agent-facing documents. Use when creating or editing skills, AGENTS.md, or CLAUDE.md.
write-adr
Write, revise, or supersede architecture decision records (ADRs) documenting an architectural choice and its rationale.
wizard
Wizard scripts for human-only provisioning, credential/CI setup, unfamiliar dashboards, migrations, or cutovers; use only for steps the agent cannot perform.
verify
Check whether the current task's claimed outcome is confirmed.
verb-tense
Choose verb tense and mood by when each event happens and how certain it is. Use when writing or reviewing any text, and when a reader asks for a tense review.